- There are plenty of laws that you may or may not be aware of whose enforcement is disparate across the population, or perhaps not enforced at all.
For example, in Texas it's actually illegal to turn without using your turn signal at least 100 feet before the turn [^1]. There are similar laws that exist in California.
If you've ever driven in Texas for even an hour within the city or suburbs you'll know this law is rarely enforced, even when police are present.
Given the lack of enforcement, the (illegal) behavior is normalized. However, despite the normalization the behavior itself is still illegal.
Now here comes the tricky part - since everyone is engaging in illegal behavior, the police, if desired, could focus on any group and trivially reprimand them for their (technically) illegal behavior. This enforcement will be reflected in the demographic likelihood of breaking this particular rule which will reinforce the very focusing on certain demographics. Recursive, if you will.
Unfortunately despite being aware of this I'm not sure what the solution would be, other than mass-surveillance. Ultimately you would need to know (1) the rate at which groups are breaking the rule absolutely, (2) the rate of which enforcement is overlooked and (3) finally the rate at which enforcement occurs. We only have but a small piece of the puzzle here.
As long as this paradox exists police will have plausible deniability backed by their very own (misleading) stats. In my opinion this is the main driver of seemingly racist law enforcement. That is, selective enforcement.
^1 - https://statutes.capitol.texas.gov/Docs/TN/htm/TN.545.htm